webAI Releases TwIL-8B, an 8.8B Reasoning Model Scoring 98.2 on MATH-500
webAI released TwIL-8B, an 8.8B reasoning model post-trained from IBM Granite 4.2 8B. It posts 98.2 on MATH-500 and 81.6 on LiveCodeBench, but trails similar-size models on knowledge benchmarks.
In brief
- What: webAI Releases TwIL-8B, an 8.8B Reasoning Model Scoring 98.2 on MATH-500
- When: November 5, 2026 at 01:25
- Source: MarkTechPost (media)
- Topic: Models
Prepared by Ilya Tarasov’s Tarasov Code newsroom based on a report by MarkTechPost.
